|
|
In this section, a comparison of three different types of models using the developer’s wood model in LS-DYNA is made using different computer platforms. The results from an Intel®-based PC (using Windows) are provided by the developer, while the results from an SGI Octane® (using UNIX) are provided by the user. Although the results are shown to be somewhat different on the different computer platforms, they are considered to be within an acceptable range based on previous experiences using different computers. This phenomenon is well known and is documented in the LS DYNA user’s manual. It is a computer platform issue and not a software issue.
The models discussed in this section are good for verifying the accuracy of the codes between computer platforms, but are shown to be unacceptable for validating the wood model itself as an accurate material for modeling wood in roadside hardware applications. However, investigating the wood material model itself is left for the remaining sections of this report.
Four single-element models were run to check the consistency between PC (Intel)-based computers and SGI-based computers. The specific results for each model are described in tables 4, 5, 6, and 7 below. In general, for single-element models, PC and SGI computers give equivalent results. Material summaries (matsum) and global statistics (glstat) were consistent throughout. Also, displacements were consistent throughout.
| Output File Examined | Variables Checked | Differences Between Computers |
|---|---|---|
| d3plot (history) | Effective stress | No visible differences |
| glstat | Internal energy, kinetic energy, and time-step size | No visible differences |
| Matsum | Internal energy and kinetic energy | No visible differences |
| Output File Examined | Variables Checked | Differences Between Computers |
|---|---|---|
| d3plot (history) | x-stress, z-stress, and pressure | No visible differences |
| Glstat | Internal energy, kinetic energy, and total energy | No visible differences |
| Output File Examined | Variables Checked | Differences Between Computers |
|---|---|---|
| d3hsp | Inertial tensor and principal directions | Slight differences for values that are near zero |
| d3plot (history) | Nodal: x, y, and z displacements Element: plastic strain and stress |
No visible differences |
| Glstat | Time-step size | No visible differences |
| Matsum | Internal energy Kinetic energy |
No visible differences Slight difference; however, magnitudes are very small (E 6) |
| Output File Examined | Variables Checked | Differences Between Computers |
|---|---|---|
| d3plot (history) | Plastic strain, stress, and pressure | No visible differences |
| Glstat | Internal energy and time-step size | No visible differences |
| Matsum | Internal energy | No visible differences |
This model is a rather detailed model of the dynamic post tests performed at the user’s facility. As evidenced in figures 62 through 67, the results from the developer’s computers match the results from the user’s computers very well for the first 15 ms of simulation. After that time, the results begin to diverge a little with regards to the contact forces and internal energy absorbed by the post (as shown in figures 64 and 66, respectively). Overall, agreement between the results is acceptable.
Model Validity: Although the developer and the user are getting nearly the same results for this model, the model itself is unacceptable for evaluating the validity of the wood model. This is because the contact between the wood and the neoprene-lined concrete sleeve is not behaving appropriately (as shown in figure 68). The interpenetration of the neoprene into the wood causes a local lockup that prevents the post from sliding along that edge.
Figure 62. Impact sequence of post simulation.
(a) Developer’s results
(b) User’s results
Figure 63. Damage (stored as effective plastic strain in d3plot files).
 Figure 64. Contact forces. |
 Figure 65. Cross section at ground level. |
 Figure 66. Energy of post parts. |
 Figure 67. Bogie velocity. |
Figure 68. Contact penetrations caused locking of parts.
The fast bogie model is a simplification of the bogie model described above. This model was generated by the developer to speed up the calculation time. As evidenced in figures 69 through 71, the results from the developer’s computers match the results from the user’s computers very well for the first 10 ms of simulation. After that time, the results begin to diverge. Overall, agreement between the results is acceptable, considering the crudeness of the model.
Model Validity: Because of the excessive bending of the post without total fracture at ground level, this model is considered to be unacceptable for evaluating the validity of the wood model.
 (a) 0 ms (developer) |
 (b) 0 ms (user) |
 (b) 10 ms (developer) |
 (b) 10 ms (user) |
 (b) 40 ms (developer) |
 (b) 40 ms (user) |
| Figure 69. Sequence of fast bogie simulations. | |
Figure 70. Energy of post parts for fast bogie simulation.
Figure 71. Section forces through post just below impact: Fast bogie simulation.
Previous | Table of Contents | Next
FHWA-HRT-04-096
